's Avatar

@blef.fr.bsky.social

I have launched Excel once

799 Followers  |  149 Following  |  47 Posts  |  Joined: 30.10.2024  |  1.9535

Latest posts by blef.fr on Bluesky

hey, thanks for the mention 😊

18.05.2025 22:58 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Geeeeeek (i love it)

25.02.2025 16:09 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

I like the idea and to be honest this is something I already had in mind (the plan exposition)β€”let’s see what it means over the weekend 😊

21.02.2025 20:54 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

oh let me have a look at this

21.02.2025 20:45 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

DuckDB team announced 2025 will be the year of the LakeHouse focus πŸ₯Ή

31.01.2025 14:17 β€” πŸ‘ 4    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

DuckCon #6 a few hours away, Pakhuis De Zwijger or live on the @duckdb.org Youtube channel.

duckdb.org/events/2025/...

Speaker are impressive, looking forward to hear from them, chance to see in person a bunch of GitHub/Discord handles / talk DuckDB a bunch.

I will be around, come to say quack!

31.01.2025 09:19 β€” πŸ‘ 16    πŸ” 5    πŸ’¬ 1    πŸ“Œ 1

Extraordinarily detailed article here by @mliebreich.bsky.social - if you're interested at all in the energy impact of AI data center buildouts I recommend spending some time with this, I learned a ton from it

12.01.2025 00:07 β€” πŸ‘ 145    πŸ” 40    πŸ’¬ 7    πŸ“Œ 2

    The GPT-4 barrier was comprehensively broken
    Some of those GPT-4 models run on my laptop
    LLM prices crashed, thanks to competition and increased efficiency
    Multimodal vision is common, audio and video are starting to emerge
    Voice and live camera mode are science fiction come to life
    Prompt driven app generation is a commodity already
    Universal access to the best models lasted for just a few short months
    β€œAgents” still haven’t really happened yet
    Evals really matter
    Apple Intelligence is bad, Apple’s MLX library is excellent
    The rise of inference-scaling β€œreasoning” models
    Was the best currently available LLM trained in China for less than $6m?
    The environmental impact got better
    The environmental impact got much, much worse
    The year of slop
    Synthetic training data works great
    LLMs somehow got even harder to use
    Knowledge is incredibly unevenly distributed
    LLMs need better criticism
    Everything tagged β€œllms” on my blog in 2024

The GPT-4 barrier was comprehensively broken Some of those GPT-4 models run on my laptop LLM prices crashed, thanks to competition and increased efficiency Multimodal vision is common, audio and video are starting to emerge Voice and live camera mode are science fiction come to life Prompt driven app generation is a commodity already Universal access to the best models lasted for just a few short months β€œAgents” still haven’t really happened yet Evals really matter Apple Intelligence is bad, Apple’s MLX library is excellent The rise of inference-scaling β€œreasoning” models Was the best currently available LLM trained in China for less than $6m? The environmental impact got better The environmental impact got much, much worse The year of slop Synthetic training data works great LLMs somehow got even harder to use Knowledge is incredibly unevenly distributed LLMs need better criticism Everything tagged β€œllms” on my blog in 2024

Here's my end-of-year review of things we learned out about LLMs in 2024 - we learned a LOT of things simonwillison.net/2024/Dec/31/...

Table of contents:

31.12.2024 18:10 β€” πŸ‘ 659    πŸ” 150    πŸ’¬ 28    πŸ“Œ 47

If Korben talks about it means its mainstream now @hannes.muehleisen.org haha

18.12.2024 20:19 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Hannes keynote at the Forward Data Conference. Really enjoyed his talk, this is for this kind of content that I’ve organised the conf πŸ”₯

18.12.2024 20:17 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

yeah

18.12.2024 20:15 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0
Preview
The State of Data Engineering in 2024: Key Insights and Trends A Look Back at the Year's Defining Patterns in Data Engineering

DEW published its yearly State of Data Engineering: Key Insights and Trends. This one-edition summary summarizes the key patterns of all DEW editions this year. Later this week, we will follow up on DEW's data engineering prediction for 2025 and beyond.

www.dataengineeringw...

16.12.2024 17:40 β€” πŸ‘ 3    πŸ” 1    πŸ’¬ 0    πŸ“Œ 0

From what I’ve seen Windsurf is great especially at frontend tasks as well

14.12.2024 17:25 β€” πŸ‘ 4    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

When simpler means that i need an AWS account?

13.12.2024 17:25 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

I see Glue as an imposter here in the sentence haha

13.12.2024 09:03 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

yes youre right if I remember correctly the discussions i had with them a few months ago

09.12.2024 20:15 β€” πŸ‘ 2    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Preview
I can now run a GPT-4 class model on my laptop Meta’s new Llama 3.3 70B is a genuinely GPT-4 class Large Language Model that runs on my laptop. Just 20 months ago I was amazed to see something that felt …

I can now run a GPT-4 class model on my laptop

(The exact same laptop that could just about run a GPT-3 class model 20 months ago)

The new Llama 3.3 70B is a striking example of the huge efficiency gains we've seen in the last two years
simonwillison.net/2024/Dec/9/l...

09.12.2024 15:19 β€” πŸ‘ 361    πŸ” 59    πŸ’¬ 11    πŸ“Œ 6
Post image

Foursquare places data is live in the hive 🐝 🍯

@hachej.bsky.social @seifert.blue

30.11.2024 00:04 β€” πŸ‘ 38    πŸ” 7    πŸ’¬ 6    πŸ“Œ 1

why are we so addicted to 3 letters when it comes to naming stuff?

This week it's MCP

01.12.2024 18:50 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

yato 🫣

01.12.2024 08:36 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

this is the exact wrap-up of the situation, but when taking a step back, imho, what we are trying to replicate with Iceberg is just what comes natively with DuckDB, so sometimes I'm like, why do we do this to ourselves

30.11.2024 12:02 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

Here in Paris its fairly common haha

29.11.2024 19:08 β€” πŸ‘ 3    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

I see DuckDB as a format yes, either standard either just metadata.

DuckDB removes the need for a specific catalog, you have it embedded directly with it

29.11.2024 12:06 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0

even more 20% / 80% haha

28.11.2024 21:05 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

that's fun you're saying this, I tend disagree, I think we can't do both because they are fundamentally different imho, I don't buy the overcomplexity of having many formats over the simplicity of picking only one

tho, I like what you're building

I have to write more on this, but for later

28.11.2024 21:04 β€” πŸ‘ 0    πŸ” 0    πŸ’¬ 2    πŸ“Œ 0

joined STATION F, now I'm a founder (lol)

28.11.2024 17:43 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

yes for sure!

27.11.2024 07:09 β€” πŸ‘ 1    πŸ” 0    πŸ’¬ 0    πŸ“Œ 0

god i feel empty, no conference to organise anymore πŸ₯Ή

yesterday was a crazy day after 6 months of work for the Forward Data Conference β€” so happy of the feedbacks and how it went

26.11.2024 11:17 β€” πŸ‘ 10    πŸ” 0    πŸ’¬ 1    πŸ“Œ 0
Post image

Salut Paris!

Forward Data Conference is amazing.

25.11.2024 11:16 β€” πŸ‘ 22    πŸ” 1    πŸ’¬ 0    πŸ“Œ 0
Post image

Awesome Keynote by @hannes.muehleisen.org about Changing Large Tables at the Forward Data Conference β€” Iceberg the Duck is coming for you

25.11.2024 11:12 β€” πŸ‘ 18    πŸ” 2    πŸ’¬ 1    πŸ“Œ 0

@blef.fr is following 20 prominent accounts